Reapplying bayesian classification

ziggy on 2003-04-08T22:11:39

Skip Montanaro describes an interesting non-email usage of spambayes: weeding out bad venues in a database of live music performances.

Hm. Perhaps there are other more interesting applications of bayesian filtering, like the venerable slashdot user commentary system....


any reason...

darobin on 2003-04-09T09:16:11

...why people are mostly using that for spam instead of providing libs to make generic automated document classification available to MUAs and the such (spam being just one category, possibly opposed to all others)?

Re:any reason...

Matts on 2003-04-09T14:35:07

What's wrong with AI::Categorize?

Re:any reason...

darobin on 2003-04-09T15:00:08

No, I know they exist, I'm just wondering why they're not being exposed in mailers and (under)used only for spam stuff. "There's no good reason" is a perfectly fine answer :)

Re:any reason...

Ovid on 2003-04-15T23:57:56

The answer's simple: because so many people scream about spam that it's a topic that's going to truly interest people. Grab any person on the street and ask them if they have a problem with spam and, if so, would they like a solution to it? We both know the likely answer :) There's going to be both a larger base of people affected and a larger percentage of them who want a solution.

Now grab any person on the street and ask them if generic automated document classification is a solution to any problem they have ...

It's not that the latter might not be a solution, but fewer people talking about it or asking for it, it's not going to be as "sexy" a problem to work on.

Re:any reason...

darobin on 2003-04-16T08:46:00

Yes, agreed, what surprises me a little is that since 99% of what it takes for document classification to happen is now already in MUAs, it ain't exposed in some remote/advanced/scary setting somewhere :)